Salisbury said he and the research team went into this project from the perspective that there were two issues to resolve.
One involved perception — how does the machine differentiate fruit from leaves — and the other revolved around getting fruit off the trees without bruising it.
The researchers had already determined that a vacuum-based picker could avoid damaging or dislodging the target fruit, adjacent fruit or part of the tree in ways that a grasping machine might.
The suction flow is intended to pull the apple to the nozzle, separate it from the tree and pass it through the nozzle to a conveyor.
The nozzles are like cartridges, coming in multiple sizes to harvest varying sizes of fruit, though the size range of a single cartridge is pretty wide, and growers are unlikely to need to swap it out within a block or even an entire orchard, Salisbury said.
Currently, the system can pitch to about 45 degrees to match the angle of a trellis and has a vertical range of about 4 to 5 feet. The vision is to have one machine picking high and another picking low from the same moving platform, Salisbury said. “We’re hoping the tractor driver is just putting it in low gear and he’s not even looking back,” he said.
Dan Steere, an SRI consultant working with Salisbury on the project, said it’s clear that growers are considering trellises because it’s by far the most efficient production system. In terms of mechanically automating harvest, a trellis system is going to be important for growers, but what kind of trellis remains to be seen, he said.
“As far as recommendations to growers to prepare for automation, we’re not in a position to offer those yet,” he told Good Fruit Grower. “It’s important for us to learn from world-class producers, world-class experts, so Washington is the ideal place to collaborate on this research.”
In 2015, the research centered on three areas:
—Refining the nozzle design to reduce stem and spur pulls and to minimize damage to the apple.
—Developing an ultra-compact decelerator to slow down the apples upon exiting the vacuum tube for conveyance.
—Integrating the picking vacuum on a commercial robot arm.
The research team integrated the vacuum picker with a commercial robot arm and a 3D stereo sensor developed by Carnegie Mellon University, showing that the end-effector design is compatible with a robot arm.
The team also determined that memory foam, with its preferred viscoelastic properties, was the ideal material for decelerating apples upon exiting the vacuum tube without bruising them.
The team also took the preliminary step of commissioning a formal experiment across multiple cultivars to determine the effect of pulled stems on fruit decay in storage. Those apples are currently in controlled-atmosphere (CA) storage and will be evaluated alongside traditionally harvested fruit beginning in March.
Abundant Robotics tested and evaluated the equipment at seven Washington orchards last fall, including McDougall and Sons, Chiawana Orchards, Yakima Valley Orchards and Matson Fruit.
As part of those demonstrations, the Washington Tree Fruit Research Commission collected 180 of the harvested apples, all Fujis, for evaluation. Seven apples were downgraded for bruising and 26 apples were culled, mostly for unanticipated punctures or cuts, Salisbury said.
“Basically, what’s causing these cuts is the vacuum system pulling in a twig or branch that then gets the apple,” he said. “Even the bruises look like it’s the same phenomenon, it just wasn’t aggressive enough to cut it.”
Salisbury said he asked growers who attended the demonstrations about it and most indicated that the problem twigs could be pruned before harvest. Of the 180 apples evaluated, 86 percent would have packed as Washington extra fancy.
Overall, the testing activities in 2015 found that the vision system and vacuum picker are capable of recognizing, localizing and picking apples without bruising them. The two systems were able to work together to pick fruit at a picking rate of faster than one apple per second.
In Australia, the researchers will be evaluating the system as a continuous picking machine, moving down the row and harvesting fruit, rather than parking it next to a tree to harvest fruit as they have in the past.
They also aim to demonstrate whether the vision system can evaluate for ripeness based on color and are researching conveyance to determine how best the apples should travel to bins from the picker.
Salisbury, who grew up in Richland, Washington, first approached the idea of developing a robotic harvester four years ago. The company aims to have the product commercially available in the fall of 2018, though Salisbury stressed that was an extremely tentative date.
“We need to make sure that limitations are set properly, that we’re prepared to deliver what we’re talking about,” he said.
Steere agreed. “A lot of people have made aggressive claims about what robotics can do in agriculture,” he said. “We don’t want to be one of the groups that claims more than we can deliver.”
That rollout — whether the equipment will be available for purchase or lease to growers, offered as part of a harvest service or some other option altogether — and the cost to growers also have not yet been determined, Salisbury said. •
– by Shannon Dininny